DOMAINcfg 1-way nesting

Versions v4.2.x
1

I use DOMAINcfg to create a nested domain, e.g. files 1_domain_cfg.nc and domain_cfg.nc
I found that while creating a nested domain DOMAINcfg also changes bathymetry in parent domain in the area where nested domain is situated.
It seems that 2-way nesting is TRUE by default in DOMAINcfg.
Am I correct?
Is there any way to turn off 2-way nesting in DOMAINcfg and use 1-way instead?

2

Yes, two-way nesting is the default with the DOMAINcfg tool, and as such, it modfies the parent mesh. We could indeed add 1-way mode as an option.
For the time being, to run in 1-way mode, simply generate a parent mesh file without any zoom (tip: set to 0 the first line of your AGRIF_FixedGrids.in file, so that you don’t have to recompile the tool without key_agrif defined).

3

Hi,

I have two questions regarding to this issue;

  1. If I run with AGRIF_FixedGrids.in, and parent grid is changed, but I do not like the topography and rerun with setting AGRIF_FixedGrids.in to zero. So I have two file 1_domain_cfg.nc and domain_cfg.nc where the latter is not adjusted according to the child grid, can I still run with 2 way nesting?
    In my case, the problem is I am trying to represent very narrow channel, and although child grid looks good, the channel gets closed in the parent grid due to parent grid modification. I do not want that.
  2. After I created my parent and child domaingrid files, I realized there is one grid wall all around the child grid’s four sides, however in the AGRIF_DEMO folder, the child grids do not contain these walls. Is this OK?

Thanks,
Mehmet

4

Hi

No, you can’t. In 2-way mode, you need to have a parent bathymetry that is exactly a volume average of corresponding child grid values. The geometry of your channel over the parent grid should not matter much (but it does a bit indeed through the prediction of the barotropic mode over the whole parent domain). The parent grid solution is anyway replaced by the child grid one.

Yes it is normal to have these masked (ghosts) points around the child domains. I wonder why this is not the case in AGRIF_DEMO.

Jérôme